generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
6c280c2299
vpx/vp8/encoder/dct.c

1110 lines
38 KiB
C
Raw Normal View History

Initial WebM release
2010-05-18 17:58:33 +02:00
/*
Use WebM in copyright notice for consistency Changes 'The VP8 project' to 'The WebM project', for consistency with other webmproject.org repositories. Fixes issue #97. Change-Id: I37c13ed5fbdb9d334ceef71c6350e9febed9bbba
2010-09-09 14:16:39 +02:00
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
Initial WebM release
2010-05-18 17:58:33 +02:00
*
cosmetics: trim trailing whitespace When the license headers were updated, they accidentally contained trailing whitespace, so unfortunately we have to touch all the files again. Change-Id: I236c05fade06589e417179c0444cb39b09e4200d
2010-06-18 18:39:21 +02:00
* Use of this source code is governed by a BSD-style license
LICENSE: update with latest text Change-Id: Ieebea089095d9073b3a94932791099f614ce120c
2010-06-04 22:19:40 +02:00
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
cosmetics: trim trailing whitespace When the license headers were updated, they accidentally contained trailing whitespace, so unfortunately we have to touch all the files again. Change-Id: I236c05fade06589e417179c0444cb39b09e4200d
2010-06-18 18:39:21 +02:00
* in the file PATENTS. All contributing project authors may
LICENSE: update with latest text Change-Id: Ieebea089095d9073b3a94932791099f614ce120c
2010-06-04 22:19:40 +02:00
* be found in the AUTHORS file in the root of the source tree.
Initial WebM release
2010-05-18 17:58:33 +02:00
*/
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
#include <assert.h>
Initial WebM release
2010-05-18 17:58:33 +02:00
#include <math.h>
experiment extending the quantizer range Prior to this change, VP8 min quantizer is 4, which caps the highest quality around 51DB. This experimental change extends the min quantizer to 1, removes the cap and allows the highest quality to be around ~73DB, consistent with the fdct/idct round trip error. To test this change, at configure time use options: --enable-experimental --enable-extend_qrange The following is a brief log of changes in each of the patch sets patch set 1: In this commit, the quantization/dequantization constants are kept unchanged, instead scaling factor 4 is rolled into fdct/idct. Fixed Q0 encoding tests on mobile: Before: 9560.567kbps Overall PSNR:50.255DB VPXSSIM:98.288 Now: 18035.774kbps Overall PSNR:73.022DB VPXSSIM:99.991 patch set 2: regenerated dc/ac quantizer lookup tables based on the scaling factor rolled in the fdct/idct. Also slightly extended the range towards the high quantizer end. patch set 3: slightly tweaked the quantizer tables and generated bits_per_mb table based on Paul's suggestions. patch set 4: fix a typo in idct, re-calculated tables relating active max Q to active min Q patch set 5: added rdmult lookup table based on Q patch set 6: fix rdmult scale: dct coefficient has scaled up by 4 patch set 7: make transform coefficients to be within 16bits patch set 8: normalize 2nd order quantizers patch set 9: fix mis-spellings patch set 10: change the configure script and macros to allow experimental code to be enabled at configure time with --enable-extend_qrange patch set 11: rebase for merge Change-Id: Ib50641ddd44aba2a52ed890222c309faa31cc59c
2010-12-02 00:50:14 +01:00
#include "vpx_ports/config.h"
Add lossless compression mode. This commit adds lossless compression capability to the experimental branch. The lossless experiment can be enabled using --enable-lossless in configure. When the experiment is enabled, the encoder will use lossless compression mode by command line option --lossless, and the decoder automatically recognizes a losslessly encoded clip and decodes accordingly. To achieve the lossless coding, this commit has changed the following: 1. To encode at lossless mode, encoder forces the use of unit quantizer, i.e, Q 0, where effective quantization is 1. Encoder also disables the usage of 8x8 transform and allows only 4x4 transform; 2. At Q 0, the first order 4x4 DCT/IDCT have been switched over to a pair of forward and inverse Walsh-Hadamard Transform (http://goo.gl/EIsfy), with proper scaling applied to match the range of the original 4x4 DCT/IDCT pair; 3. At Q 0, the second order remains to use the previous walsh-hadamard transform pair. However, to maintain the reversibility in second order transform at Q 0, scaling down is applied to first order DC coefficients prior to forward transform, and scaling up is applied to the second order output prior to quantization. Symmetric upscaling and downscaling are added around inverse second order transform; 4. At lossless mode, encoder also disables a number of minor features to ensure no loss is introduced, these features includes: a. Trellis quantization optimization b. Loop filtering c. Aggressive zero-binning, rounding and zero-bin boosting d. Mode based zero-bin boosting Lossless coding test was performed on all clips within the derf set, to verify that the commit has achieved lossless compression for all clips. The average compression ratio is around 2.57 to 1. (http://goo.gl/dEShs) Change-Id: Ia3aba7dd09df40dd590f93b9aba134defbc64e34
2012-06-14 04:03:31 +02:00
#include "vp8/common/idct.h"
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
#include "vp8/common/systemdependent.h"
Add 8x8 transform to experimental branch Please refer to previous commit messages for detailed info: https://on2-git.corp.google.com/g/#change,5940 https://on2-git.corp.google.com/g/#change,6045 Change-Id: I8b16992f2f69c5a808ad40a3e32ef589cce7c59d
2011-02-14 23:18:18 +01:00
Adds hybrid transform Adds ADST/DCT hybrid transform coding for Intra4x4 mode. The ADST is applied to directions in which the boundary pixels are used for prediction, while DCT applied to directions without corresponding boundary prediction. Adds enum TX_TYPE in b_mode_infor to indicate the transform type used. Make coding style consistent with google style. Fixed the commented issues. Experimental results in terms of bit-rate reduction: derf: 0.731% yt: 0.982% std-hd: 0.459% hd: 0.725% Will be looking at 8x8 transforms next. Change-Id: I46dbd7b80dbb3e8856e9c34fbc58cb3764a12fcf
2012-06-25 21:26:09 +02:00
#include "vp8/common/blockd.h"
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
// TODO: these transforms can be converted into integer forms to reduce
// the complexity
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const float dct_4[16] = {
Adds hybrid transform Adds ADST/DCT hybrid transform coding for Intra4x4 mode. The ADST is applied to directions in which the boundary pixels are used for prediction, while DCT applied to directions without corresponding boundary prediction. Adds enum TX_TYPE in b_mode_infor to indicate the transform type used. Make coding style consistent with google style. Fixed the commented issues. Experimental results in terms of bit-rate reduction: derf: 0.731% yt: 0.982% std-hd: 0.459% hd: 0.725% Will be looking at 8x8 transforms next. Change-Id: I46dbd7b80dbb3e8856e9c34fbc58cb3764a12fcf
2012-06-25 21:26:09 +02:00
0.500000000000000, 0.500000000000000, 0.500000000000000, 0.500000000000000,
0.653281482438188, 0.270598050073099, -0.270598050073099, -0.653281482438188,
0.500000000000000, -0.500000000000000, -0.500000000000000, 0.500000000000000,
0.270598050073099, -0.653281482438188, 0.653281482438188, -0.270598050073099
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const float adst_4[16] = {
Adds hybrid transform Adds ADST/DCT hybrid transform coding for Intra4x4 mode. The ADST is applied to directions in which the boundary pixels are used for prediction, while DCT applied to directions without corresponding boundary prediction. Adds enum TX_TYPE in b_mode_infor to indicate the transform type used. Make coding style consistent with google style. Fixed the commented issues. Experimental results in terms of bit-rate reduction: derf: 0.731% yt: 0.982% std-hd: 0.459% hd: 0.725% Will be looking at 8x8 transforms next. Change-Id: I46dbd7b80dbb3e8856e9c34fbc58cb3764a12fcf
2012-06-25 21:26:09 +02:00
0.228013428883779, 0.428525073124360, 0.577350269189626, 0.656538502008139,
0.577350269189626, 0.577350269189626, 0.000000000000000, -0.577350269189626,
0.656538502008139, -0.228013428883779, -0.577350269189626, 0.428525073124359,
0.428525073124360, -0.656538502008139, 0.577350269189626, -0.228013428883779
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const float dct_8[64] = {
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
0.353553390593274, 0.353553390593274, 0.353553390593274, 0.353553390593274,
0.353553390593274, 0.353553390593274, 0.353553390593274, 0.353553390593274,
0.490392640201615, 0.415734806151273, 0.277785116509801, 0.097545161008064,
-0.097545161008064, -0.277785116509801, -0.415734806151273, -0.490392640201615,
0.461939766255643, 0.191341716182545, -0.191341716182545, -0.461939766255643,
-0.461939766255643, -0.191341716182545, 0.191341716182545, 0.461939766255643,
0.415734806151273, -0.097545161008064, -0.490392640201615, -0.277785116509801,
0.277785116509801, 0.490392640201615, 0.097545161008064, -0.415734806151273,
0.353553390593274, -0.353553390593274, -0.353553390593274, 0.353553390593274,
0.353553390593274, -0.353553390593274, -0.353553390593274, 0.353553390593274,
0.277785116509801, -0.490392640201615, 0.097545161008064, 0.415734806151273,
-0.415734806151273, -0.097545161008064, 0.490392640201615, -0.277785116509801,
0.191341716182545, -0.461939766255643, 0.461939766255643, -0.191341716182545,
-0.191341716182545, 0.461939766255643, -0.461939766255643, 0.191341716182545,
0.097545161008064, -0.277785116509801, 0.415734806151273, -0.490392640201615,
0.490392640201615, -0.415734806151273, 0.277785116509801, -0.097545161008064
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const float adst_8[64] = {
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
0.089131608307533, 0.175227946595735, 0.255357107325376, 0.326790388032145,
0.387095214016349, 0.434217976756762, 0.466553967085785, 0.483002021635509,
0.255357107325376, 0.434217976756762, 0.483002021635509, 0.387095214016349,
0.175227946595735, -0.089131608307533, -0.326790388032145, -0.466553967085785,
0.387095214016349, 0.466553967085785, 0.175227946595735, -0.255357107325376,
-0.483002021635509, -0.326790388032145, 0.089131608307533, 0.434217976756762,
0.466553967085785, 0.255357107325376, -0.326790388032145, -0.434217976756762,
0.089131608307533, 0.483002021635509, 0.175227946595735, -0.387095214016348,
0.483002021635509, -0.089131608307533, -0.466553967085785, 0.175227946595735,
0.434217976756762, -0.255357107325376, -0.387095214016348, 0.326790388032145,
0.434217976756762, -0.387095214016348, -0.089131608307533, 0.466553967085786,
-0.326790388032145, -0.175227946595735, 0.483002021635509, -0.255357107325375,
0.326790388032145, -0.483002021635509, 0.387095214016349, -0.089131608307534,
-0.255357107325377, 0.466553967085785, -0.434217976756762, 0.175227946595736,
0.175227946595735, -0.326790388032145, 0.434217976756762, -0.483002021635509,
0.466553967085785, -0.387095214016348, 0.255357107325376, -0.089131608307532
};
Adds hybrid transform Adds ADST/DCT hybrid transform coding for Intra4x4 mode. The ADST is applied to directions in which the boundary pixels are used for prediction, while DCT applied to directions without corresponding boundary prediction. Adds enum TX_TYPE in b_mode_infor to indicate the transform type used. Make coding style consistent with google style. Fixed the commented issues. Experimental results in terms of bit-rate reduction: derf: 0.731% yt: 0.982% std-hd: 0.459% hd: 0.725% Will be looking at 8x8 transforms next. Change-Id: I46dbd7b80dbb3e8856e9c34fbc58cb3764a12fcf
2012-06-25 21:26:09 +02:00
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
/* Converted the transforms to integers. */
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const int16_t dct_i4[16] = {
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
16384, 16384, 16384, 16384,
21407, 8867, -8867, -21407,
16384, -16384, -16384, 16384,
8867, -21407, 21407, -8867
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const int16_t adst_i4[16] = {
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
7472, 14042, 18919, 21513,
18919, 18919, 0, -18919,
21513, -7472, -18919, 14042,
14042, -21513, 18919, -7472
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const int16_t dct_i8[64] = {
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
11585, 11585, 11585, 11585,
11585, 11585, 11585, 11585,
16069, 13623, 9102, 3196,
-3196, -9102, -13623, -16069,
15137, 6270, -6270, -15137,
-15137, -6270, 6270, 15137,
13623, -3196, -16069, -9102,
9102, 16069, 3196, -13623,
11585, -11585, -11585, 11585,
11585, -11585, -11585, 11585,
9102, -16069, 3196, 13623,
-13623, -3196, 16069, -9102,
6270, -15137, 15137, -6270,
-6270, 15137, -15137, 6270,
3196, -9102, 13623, -16069,
16069, -13623, 9102, -3196
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const int16_t adst_i8[64] = {
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
2921, 5742, 8368, 10708,
12684, 14228, 15288, 15827,
8368, 14228, 15827, 12684,
5742, -2921, -10708, -15288,
12684, 15288, 5742, -8368,
-15827, -10708, 2921, 14228,
15288, 8368, -10708, -14228,
2921, 15827, 5742, -12684,
15827, -2921, -15288, 5742,
14228, -8368, -12684, 10708,
14228, -12684, -2921, 15288,
-10708, -5742, 15827, -8368,
10708, -15827, 12684, -2921,
-8368, 15288, -14228, 5742,
5742, -10708, 14228, -15827,
15288, -12684, 8368, -2921
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const float dct_16[256] = {
hybrid transform of 16x16 dimension Enable ADST/DCT of dimension 16x16 for I16X16 modes. This change provides benefits mostly for hd sequences. Set up the framework for selectable transform dimension. Also allowing quantization parameter threshold to control the use of hybrid transform (This is currently disabled by setting threshold always above the quantization parameter. Adaptive thresholding can be built upon this, which will further improve the coding performance.) The coding performance gains (with respect to the codec that has all other configuration settings turned on) are derf: 0.013 yt: 0.086 hd: 0.198 std-hd: 0.501 Change-Id: Ibb4263a61fc74e0b3c345f54d73e8c73552bf926
2012-08-29 20:25:38 +02:00
0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000,
0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000, 0.250000,
0.351851, 0.338330, 0.311806, 0.273300, 0.224292, 0.166664, 0.102631, 0.034654,
-0.034654, -0.102631, -0.166664, -0.224292, -0.273300, -0.311806, -0.338330, -0.351851,
0.346760, 0.293969, 0.196424, 0.068975, -0.068975, -0.196424, -0.293969, -0.346760,
-0.346760, -0.293969, -0.196424, -0.068975, 0.068975, 0.196424, 0.293969, 0.346760,
0.338330, 0.224292, 0.034654, -0.166664, -0.311806, -0.351851, -0.273300, -0.102631,
0.102631, 0.273300, 0.351851, 0.311806, 0.166664, -0.034654, -0.224292, -0.338330,
0.326641, 0.135299, -0.135299, -0.326641, -0.326641, -0.135299, 0.135299, 0.326641,
0.326641, 0.135299, -0.135299, -0.326641, -0.326641, -0.135299, 0.135299, 0.326641,
0.311806, 0.034654, -0.273300, -0.338330, -0.102631, 0.224292, 0.351851, 0.166664,
-0.166664, -0.351851, -0.224292, 0.102631, 0.338330, 0.273300, -0.034654, -0.311806,
0.293969, -0.068975, -0.346760, -0.196424, 0.196424, 0.346760, 0.068975, -0.293969,
-0.293969, 0.068975, 0.346760, 0.196424, -0.196424, -0.346760, -0.068975, 0.293969,
0.273300, -0.166664, -0.338330, 0.034654, 0.351851, 0.102631, -0.311806, -0.224292,
0.224292, 0.311806, -0.102631, -0.351851, -0.034654, 0.338330, 0.166664, -0.273300,
0.250000, -0.250000, -0.250000, 0.250000, 0.250000, -0.250000, -0.250000, 0.250000,
0.250000, -0.250000, -0.250000, 0.250000, 0.250000, -0.250000, -0.250000, 0.250000,
0.224292, -0.311806, -0.102631, 0.351851, -0.034654, -0.338330, 0.166664, 0.273300,
-0.273300, -0.166664, 0.338330, 0.034654, -0.351851, 0.102631, 0.311806, -0.224292,
0.196424, -0.346760, 0.068975, 0.293969, -0.293969, -0.068975, 0.346760, -0.196424,
-0.196424, 0.346760, -0.068975, -0.293969, 0.293969, 0.068975, -0.346760, 0.196424,
0.166664, -0.351851, 0.224292, 0.102631, -0.338330, 0.273300, 0.034654, -0.311806,
0.311806, -0.034654, -0.273300, 0.338330, -0.102631, -0.224292, 0.351851, -0.166664,
0.135299, -0.326641, 0.326641, -0.135299, -0.135299, 0.326641, -0.326641, 0.135299,
0.135299, -0.326641, 0.326641, -0.135299, -0.135299, 0.326641, -0.326641, 0.135299,
0.102631, -0.273300, 0.351851, -0.311806, 0.166664, 0.034654, -0.224292, 0.338330,
-0.338330, 0.224292, -0.034654, -0.166664, 0.311806, -0.351851, 0.273300, -0.102631,
0.068975, -0.196424, 0.293969, -0.346760, 0.346760, -0.293969, 0.196424, -0.068975,
-0.068975, 0.196424, -0.293969, 0.346760, -0.346760, 0.293969, -0.196424, 0.068975,
0.034654, -0.102631, 0.166664, -0.224292, 0.273300, -0.311806, 0.338330, -0.351851,
0.351851, -0.338330, 0.311806, -0.273300, 0.224292, -0.166664, 0.102631, -0.034654
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const float adst_16[256] = {
hybrid transform of 16x16 dimension Enable ADST/DCT of dimension 16x16 for I16X16 modes. This change provides benefits mostly for hd sequences. Set up the framework for selectable transform dimension. Also allowing quantization parameter threshold to control the use of hybrid transform (This is currently disabled by setting threshold always above the quantization parameter. Adaptive thresholding can be built upon this, which will further improve the coding performance.) The coding performance gains (with respect to the codec that has all other configuration settings turned on) are derf: 0.013 yt: 0.086 hd: 0.198 std-hd: 0.501 Change-Id: Ibb4263a61fc74e0b3c345f54d73e8c73552bf926
2012-08-29 20:25:38 +02:00
0.033094, 0.065889, 0.098087, 0.129396, 0.159534, 0.188227, 0.215215, 0.240255,
0.263118, 0.283599, 0.301511, 0.316693, 0.329007, 0.338341, 0.344612, 0.347761,
0.098087, 0.188227, 0.263118, 0.316693, 0.344612, 0.344612, 0.316693, 0.263118,
0.188227, 0.098087, 0.000000, -0.098087, -0.188227, -0.263118, -0.316693, -0.344612,
0.159534, 0.283599, 0.344612, 0.329007, 0.240255, 0.098087, -0.065889, -0.215215,
-0.316693, -0.347761, -0.301511, -0.188227, -0.033094, 0.129396, 0.263118, 0.338341,
0.215215, 0.338341, 0.316693, 0.159534, -0.065889, -0.263118, -0.347761, -0.283599,
-0.098087, 0.129396, 0.301511, 0.344612, 0.240255, 0.033094, -0.188227, -0.329007,
0.263118, 0.344612, 0.188227, -0.098087, -0.316693, -0.316693, -0.098087, 0.188227,
0.344612, 0.263118, 0.000000, -0.263118, -0.344612, -0.188227, 0.098087, 0.316693,
0.301511, 0.301511, 0.000000, -0.301511, -0.301511, -0.000000, 0.301511, 0.301511,
0.000000, -0.301511, -0.301511, -0.000000, 0.301511, 0.301511, 0.000000, -0.301511,
0.329007, 0.215215, -0.188227, -0.338341, -0.033094, 0.316693, 0.240255, -0.159534,
-0.344612, -0.065889, 0.301511, 0.263118, -0.129396, -0.347761, -0.098087, 0.283599,
0.344612, 0.098087, -0.316693, -0.188227, 0.263118, 0.263118, -0.188227, -0.316693,
0.098087, 0.344612, 0.000000, -0.344612, -0.098087, 0.316693, 0.188227, -0.263118,
0.347761, -0.033094, -0.344612, 0.065889, 0.338341, -0.098087, -0.329007, 0.129396,
0.316693, -0.159534, -0.301511, 0.188227, 0.283599, -0.215215, -0.263118, 0.240255,
0.338341, -0.159534, -0.263118, 0.283599, 0.129396, -0.344612, 0.033094, 0.329007,
-0.188227, -0.240255, 0.301511, 0.098087, -0.347761, 0.065889, 0.316693, -0.215215,
0.316693, -0.263118, -0.098087, 0.344612, -0.188227, -0.188227, 0.344612, -0.098087,
-0.263118, 0.316693, 0.000000, -0.316693, 0.263118, 0.098087, -0.344612, 0.188227,
0.283599, -0.329007, 0.098087, 0.215215, -0.347761, 0.188227, 0.129396, -0.338341,
0.263118, 0.033094, -0.301511, 0.316693, -0.065889, -0.240255, 0.344612, -0.159534,
0.240255, -0.347761, 0.263118, -0.033094, -0.215215, 0.344612, -0.283599, 0.065889,
0.188227, -0.338341, 0.301511, -0.098087, -0.159534, 0.329007, -0.316693, 0.129396,
0.188227, -0.316693, 0.344612, -0.263118, 0.098087, 0.098087, -0.263118, 0.344612,
-0.316693, 0.188227, 0.000000, -0.188227, 0.316693, -0.344612, 0.263118, -0.098087,
0.129396, -0.240255, 0.316693, -0.347761, 0.329007, -0.263118, 0.159534, -0.033094,
-0.098087, 0.215215, -0.301511, 0.344612, -0.338341, 0.283599, -0.188227, 0.065889,
0.065889, -0.129396, 0.188227, -0.240255, 0.283599, -0.316693, 0.338341, -0.347761,
0.344612, -0.329007, 0.301511, -0.263118, 0.215215, -0.159534, 0.098087, -0.033094
};
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
/* Converted the transforms to integers. */
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const int16_t dct_i16[256] = {
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192,
8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192,
11529, 11086, 10217, 8955, 7350, 5461, 3363, 1136,
-1136, -3363, -5461, -7350, -8955, -10217, -11086, -11529,
11363, 9633, 6436, 2260, -2260, -6436, -9633, -11363,
-11363, -9633, -6436, -2260, 2260, 6436, 9633, 11363,
11086, 7350, 1136, -5461, -10217, -11529, -8955, -3363,
3363, 8955, 11529, 10217, 5461, -1136, -7350, -11086,
10703, 4433, -4433, -10703, -10703, -4433, 4433, 10703,
10703, 4433, -4433, -10703, -10703, -4433, 4433, 10703,
10217, 1136, -8955, -11086, -3363, 7350, 11529, 5461,
-5461, -11529, -7350, 3363, 11086, 8955, -1136, -10217,
9633, -2260, -11363, -6436, 6436, 11363, 2260, -9633,
-9633, 2260, 11363, 6436, -6436, -11363, -2260, 9633,
8955, -5461, -11086, 1136, 11529, 3363, -10217, -7350,
7350, 10217, -3363, -11529, -1136, 11086, 5461, -8955,
8192, -8192, -8192, 8192, 8192, -8192, -8192, 8192,
8192, -8192, -8192, 8192, 8192, -8192, -8192, 8192,
7350, -10217, -3363, 11529, -1136, -11086, 5461, 8955,
-8955, -5461, 11086, 1136, -11529, 3363, 10217, -7350,
6436, -11363, 2260, 9633, -9633, -2260, 11363, -6436,
-6436, 11363, -2260, -9633, 9633, 2260, -11363, 6436,
5461, -11529, 7350, 3363, -11086, 8955, 1136, -10217,
10217, -1136, -8955, 11086, -3363, -7350, 11529, -5461,
4433, -10703, 10703, -4433, -4433, 10703, -10703, 4433,
4433, -10703, 10703, -4433, -4433, 10703, -10703, 4433,
3363, -8955, 11529, -10217, 5461, 1136, -7350, 11086,
-11086, 7350, -1136, -5461, 10217, -11529, 8955, -3363,
2260, -6436, 9633, -11363, 11363, -9633, 6436, -2260,
-2260, 6436, -9633, 11363, -11363, 9633, -6436, 2260,
1136, -3363, 5461, -7350, 8955, -10217, 11086, -11529,
11529, -11086, 10217, -8955, 7350, -5461, 3363, -1136
};
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
static const int16_t adst_i16[256] = {
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
1084, 2159, 3214, 4240, 5228, 6168, 7052, 7873,
8622, 9293, 9880, 10377, 10781, 11087, 11292, 11395,
3214, 6168, 8622, 10377, 11292, 11292, 10377, 8622,
6168, 3214, 0, -3214, -6168, -8622, -10377, -11292,
5228, 9293, 11292, 10781, 7873, 3214, -2159, -7052,
-10377, -11395, -9880, -6168, -1084, 4240, 8622, 11087,
7052, 11087, 10377, 5228, -2159, -8622, -11395, -9293,
-3214, 4240, 9880, 11292, 7873, 1084, -6168, -10781,
8622, 11292, 6168, -3214, -10377, -10377, -3214, 6168,
11292, 8622, 0, -8622, -11292, -6168, 3214, 10377,
9880, 9880, 0, -9880, -9880, 0, 9880, 9880,
0, -9880, -9880, 0, 9880, 9880, 0, -9880,
10781, 7052, -6168, -11087, -1084, 10377, 7873, -5228,
-11292, -2159, 9880, 8622, -4240, -11395, -3214, 9293,
11292, 3214, -10377, -6168, 8622, 8622, -6168, -10377,
3214, 11292, 0, -11292, -3214, 10377, 6168, -8622,
11395, -1084, -11292, 2159, 11087, -3214, -10781, 4240,
10377, -5228, -9880, 6168, 9293, -7052, -8622, 7873,
11087, -5228, -8622, 9293, 4240, -11292, 1084, 10781,
-6168, -7873, 9880, 3214, -11395, 2159, 10377, -7052,
10377, -8622, -3214, 11292, -6168, -6168, 11292, -3214,
-8622, 10377, 0, -10377, 8622, 3214, -11292, 6168,
9293, -10781, 3214, 7052, -11395, 6168, 4240, -11087,
8622, 1084, -9880, 10377, -2159, -7873, 11292, -5228,
7873, -11395, 8622, -1084, -7052, 11292, -9293, 2159,
6168, -11087, 9880, -3214, -5228, 10781, -10377, 4240,
6168, -10377, 11292, -8622, 3214, 3214, -8622, 11292,
-10377, 6168, 0, -6168, 10377, -11292, 8622, -3214,
4240, -7873, 10377, -11395, 10781, -8622, 5228, -1084,
-3214, 7052, -9880, 11292, -11087, 9293, -6168, 2159,
2159, -4240, 6168, -7873, 9293, -10377, 11087, -11395,
11292, -10781, 9880, -8622, 7052, -5228, 3214, -1084
};
Changed to use integer 8x8 dct The commit added an integer version of 8x8 forward DCT, based on the orginal forward DCT from VP6. The constants, roundings, and shifts were adjusted to improve the accuracy. The latest patch has a very similar accuracy in term of round trip error against the floating point version. It should be noted here that the purpose of the patch is to help encoding speed and facilitate all other experiments. There will be futher review in combination with inverse DCT before finalization. configure with "--enable--int_8x8fdct" to use the integer version Change-Id: I5a4f80507429f0e07cf02a13768ec81cbfddc5bc
2012-05-09 18:31:14 +02:00
static const int xC1S7 = 16069;
static const int xC2S6 = 15137;
static const int xC3S5 = 13623;
static const int xC4S4 = 11585;
static const int xC5S3 = 9102;
static const int xC6S2 = 6270;
static const int xC7S1 = 3196;
#define SHIFT_BITS 14
#define DOROUND(X) X += (1<<(SHIFT_BITS-1));
#define FINAL_SHIFT 3
#define FINAL_ROUNDING (1<<(FINAL_SHIFT -1))
#define IN_SHIFT (FINAL_SHIFT+1)
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_short_fdct8x8_c(short *InputData, short *OutputData, int pitch) {
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
int loop;
int short_pitch = pitch >> 1;
int is07, is12, is34, is56;
int is0734, is1256;
int id07, id12, id34, id56;
int irot_input_x, irot_input_y;
int icommon_product1; // Re-used product (c4s4 * (s12 - s56))
int icommon_product2; // Re-used product (c4s4 * (d12 + d56))
int temp1, temp2; // intermediate variable for computation
int InterData[64];
int *ip = InterData;
short *op = OutputData;
for (loop = 0; loop < 8; loop++) {
// Pre calculate some common sums and differences.
is07 = (InputData[0] + InputData[7]) << IN_SHIFT;
is12 = (InputData[1] + InputData[2]) << IN_SHIFT;
is34 = (InputData[3] + InputData[4]) << IN_SHIFT;
is56 = (InputData[5] + InputData[6]) << IN_SHIFT;
id07 = (InputData[0] - InputData[7]) << IN_SHIFT;
id12 = (InputData[1] - InputData[2]) << IN_SHIFT;
id34 = (InputData[3] - InputData[4]) << IN_SHIFT;
id56 = (InputData[5] - InputData[6]) << IN_SHIFT;
is0734 = is07 + is34;
is1256 = is12 + is56;
// Pre-Calculate some common product terms.
icommon_product1 = xC4S4 * (is12 - is56);
DOROUND(icommon_product1)
icommon_product1 >>= SHIFT_BITS;
icommon_product2 = xC4S4 * (id12 + id56);
DOROUND(icommon_product2)
icommon_product2 >>= SHIFT_BITS;
ip[0] = (xC4S4 * (is0734 + is1256));
DOROUND(ip[0]);
ip[0] >>= SHIFT_BITS;
ip[4] = (xC4S4 * (is0734 - is1256));
DOROUND(ip[4]);
ip[4] >>= SHIFT_BITS;
// Define inputs to rotation for outputs 2 and 6
irot_input_x = id12 - id56;
irot_input_y = is07 - is34;
// Apply rotation for outputs 2 and 6.
temp1 = xC6S2 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC2S6 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
ip[2] = temp1 + temp2;
temp1 = xC6S2 * irot_input_y;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC2S6 * irot_input_x;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
ip[6] = temp1 - temp2;
// Define inputs to rotation for outputs 1 and 7
irot_input_x = icommon_product1 + id07;
irot_input_y = -(id34 + icommon_product2);
// Apply rotation for outputs 1 and 7.
temp1 = xC1S7 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC7S1 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
ip[1] = temp1 - temp2;
temp1 = xC7S1 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC1S7 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
ip[7] = temp1 + temp2;
// Define inputs to rotation for outputs 3 and 5
irot_input_x = id07 - icommon_product1;
irot_input_y = id34 - icommon_product2;
// Apply rotation for outputs 3 and 5.
temp1 = xC3S5 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC5S3 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
ip[3] = temp1 - temp2;
temp1 = xC5S3 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC3S5 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
ip[5] = temp1 + temp2;
// Increment data pointer for next row
InputData += short_pitch;
ip += 8;
}
Changed to use integer 8x8 dct The commit added an integer version of 8x8 forward DCT, based on the orginal forward DCT from VP6. The constants, roundings, and shifts were adjusted to improve the accuracy. The latest patch has a very similar accuracy in term of round trip error against the floating point version. It should be noted here that the purpose of the patch is to help encoding speed and facilitate all other experiments. There will be futher review in combination with inverse DCT before finalization. configure with "--enable--int_8x8fdct" to use the integer version Change-Id: I5a4f80507429f0e07cf02a13768ec81cbfddc5bc
2012-05-09 18:31:14 +02:00
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
// Performed DCT on rows, now transform the columns
ip = InterData;
for (loop = 0; loop < 8; loop++) {
// Pre calculate some common sums and differences.
is07 = ip[0 * 8] + ip[7 * 8];
is12 = ip[1 * 8] + ip[2 * 8];
is34 = ip[3 * 8] + ip[4 * 8];
is56 = ip[5 * 8] + ip[6 * 8];
id07 = ip[0 * 8] - ip[7 * 8];
id12 = ip[1 * 8] - ip[2 * 8];
id34 = ip[3 * 8] - ip[4 * 8];
id56 = ip[5 * 8] - ip[6 * 8];
is0734 = is07 + is34;
is1256 = is12 + is56;
// Pre-Calculate some common product terms
icommon_product1 = xC4S4 * (is12 - is56);
icommon_product2 = xC4S4 * (id12 + id56);
DOROUND(icommon_product1)
DOROUND(icommon_product2)
icommon_product1 >>= SHIFT_BITS;
icommon_product2 >>= SHIFT_BITS;
temp1 = xC4S4 * (is0734 + is1256);
temp2 = xC4S4 * (is0734 - is1256);
DOROUND(temp1);
DOROUND(temp2);
temp1 >>= SHIFT_BITS;
temp2 >>= SHIFT_BITS;
op[0 * 8] = (temp1 + FINAL_ROUNDING) >> FINAL_SHIFT;
op[4 * 8] = (temp2 + FINAL_ROUNDING) >> FINAL_SHIFT;
// Define inputs to rotation for outputs 2 and 6
irot_input_x = id12 - id56;
irot_input_y = is07 - is34;
// Apply rotation for outputs 2 and 6.
temp1 = xC6S2 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC2S6 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
op[2 * 8] = (temp1 + temp2 + FINAL_ROUNDING) >> FINAL_SHIFT;
temp1 = xC6S2 * irot_input_y;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC2S6 * irot_input_x;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
op[6 * 8] = (temp1 - temp2 + FINAL_ROUNDING) >> FINAL_SHIFT;
// Define inputs to rotation for outputs 1 and 7
irot_input_x = icommon_product1 + id07;
irot_input_y = -(id34 + icommon_product2);
// Apply rotation for outputs 1 and 7.
temp1 = xC1S7 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC7S1 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
op[1 * 8] = (temp1 - temp2 + FINAL_ROUNDING) >> FINAL_SHIFT;
temp1 = xC7S1 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC1S7 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
op[7 * 8] = (temp1 + temp2 + FINAL_ROUNDING) >> FINAL_SHIFT;
// Define inputs to rotation for outputs 3 and 5
irot_input_x = id07 - icommon_product1;
irot_input_y = id34 - icommon_product2;
// Apply rotation for outputs 3 and 5.
temp1 = xC3S5 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC5S3 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
op[3 * 8] = (temp1 - temp2 + FINAL_ROUNDING) >> FINAL_SHIFT;
temp1 = xC5S3 * irot_input_x;
DOROUND(temp1);
temp1 >>= SHIFT_BITS;
temp2 = xC3S5 * irot_input_y;
DOROUND(temp2);
temp2 >>= SHIFT_BITS;
op[5 * 8] = (temp1 + temp2 + FINAL_ROUNDING) >> FINAL_SHIFT;
// Increment data pointer for next column.
ip++;
op++;
}
Changed to use integer 8x8 dct The commit added an integer version of 8x8 forward DCT, based on the orginal forward DCT from VP6. The constants, roundings, and shifts were adjusted to improve the accuracy. The latest patch has a very similar accuracy in term of round trip error against the floating point version. It should be noted here that the purpose of the patch is to help encoding speed and facilitate all other experiments. There will be futher review in combination with inverse DCT before finalization. configure with "--enable--int_8x8fdct" to use the integer version Change-Id: I5a4f80507429f0e07cf02a13768ec81cbfddc5bc
2012-05-09 18:31:14 +02:00
}
Add 8x8 transform to experimental branch Please refer to previous commit messages for detailed info: https://on2-git.corp.google.com/g/#change,5940 https://on2-git.corp.google.com/g/#change,6045 Change-Id: I8b16992f2f69c5a808ad40a3e32ef589cce7c59d
2011-02-14 23:18:18 +01:00
Adjust style to match Google Coding Style a little more closely. Most of these were picked up in the previous commit (prefix change from vp8_ to vp9_), but I'm pushing this separately so that it's easier to review. Change-Id: I91e959895778b8632d7d33375523df8a7568a490
2012-10-30 22:25:33 +01:00
void vp9_short_fhaar2x2_c(short *input, short *output, int pitch) {
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
/* [1 1; 1 -1] orthogonal transform */
/* use position: 0,1, 4, 8 */
int i;
short *ip1 = input;
short *op1 = output;
for (i = 0; i < 16; i++) {
op1[i] = 0;
}
op1[0] = (ip1[0] + ip1[1] + ip1[4] + ip1[8] + 1) >> 1;
op1[1] = (ip1[0] - ip1[1] + ip1[4] - ip1[8]) >> 1;
op1[4] = (ip1[0] + ip1[1] - ip1[4] - ip1[8]) >> 1;
op1[8] = (ip1[0] - ip1[1] - ip1[4] + ip1[8]) >> 1;
Add 8x8 transform to experimental branch Please refer to previous commit messages for detailed info: https://on2-git.corp.google.com/g/#change,5940 https://on2-git.corp.google.com/g/#change,6045 Change-Id: I8b16992f2f69c5a808ad40a3e32ef589cce7c59d
2011-02-14 23:18:18 +01:00
}
Merge t8x8 experiments Change-Id: I8e9b6b154e1a0d0cb42d596366380d69c00ac15f
2012-02-29 02:11:12 +01:00
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
/* For test */
#define TEST_INT 1
#if TEST_INT
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
#define vp9_fht_int_c vp9_fht_c
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
#else
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
#define vp9_fht_float_c vp9_fht_c
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
#endif
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_fht_float_c(const int16_t *input, int pitch, int16_t *output,
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
TX_TYPE tx_type, int tx_dim) {
Remove vp8 in local symbols. For non-static functions, change the prefix to vp9_. For static functions, remove the prefix. Also fix some comments, remove unused code or unused function prototypes. Change-Id: I1f8be05362f66060fe421c3d4c9a906fdf835de5
2012-10-31 22:40:53 +01:00
vp9_clear_system_state(); // Make it simd safe : __asm emms;
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
{
int i, j, k;
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
float bufa[256], bufb[256]; // buffers are for floating-point test purpose
// the implementation could be simplified in
// conjunction with integer transform
const int16_t *ip = input;
int16_t *op = output;
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
float *pfa = &bufa[0];
float *pfb = &bufb[0];
// pointers to vertical and horizontal transforms
Change non-function symbol vp8_ prefixes to vp9_. For local symbols, make them static instead. Change-Id: I13d60947a46f711bc8991e16100cea2a13e3a22e
2012-10-31 01:12:12 +01:00
const float *ptv, *pth;
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
Some cleanups and fixes. Separates the logic on transform type selection previously spread out over a number of files into a separate function. Currently the tx_type field in b_mode_info is not used, but still left in there to eventually use for signaling the transform type in the bitstream. Also, now for tx_type = DCT_DCT, the regular integer DCT is used, as opposed to the floating point DCT used in conjuction with hybrid transform. Results change somewhat due to the transform change, but are within reasonable limits. The hd/std-hd sets are slightly up, while derf/yt are slightly down. Change-Id: I5776840c2239ca2da31ca6cfd7fd1148dc5f9e0f
2012-10-16 01:41:41 +02:00
assert(tx_type != DCT_DCT);
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
// load and convert residual array into floating-point
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
for (j = 0; j < tx_dim; j++) {
for (i = 0; i < tx_dim; i++) {
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfa[i] = (float)ip[i];
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
}
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfa += tx_dim;
ip += pitch / 2;
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
}
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
// vertical transformation
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
pfa = &bufa[0];
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfb = &bufb[0];
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
switch (tx_type) {
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
case ADST_ADST :
case ADST_DCT :
ptv = (tx_dim == 4) ? &adst_4[0] :
((tx_dim == 8) ? &adst_8[0] : &adst_16[0]);
break;
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
default :
ptv = (tx_dim == 4) ? &dct_4[0] :
((tx_dim == 8) ? &dct_8[0] : &dct_16[0]);
break;
}
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
for (j = 0; j < tx_dim; j++) {
for (i = 0; i < tx_dim; i++) {
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfb[i] = 0;
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
for (k = 0; k < tx_dim; k++) {
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfb[i] += ptv[k] * pfa[(k * tx_dim)];
}
pfa += 1;
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
}
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfb += tx_dim;
ptv += tx_dim;
pfa = &bufa[0];
Entropy coding for hybrid transform Separates the entropy coding context models for 4x4, 8x8 and 16x16 ADST variants. There is a small improvement for HD (hd/std-hd) by about 0.1-0.2%. Results on derf/yt are about the same, probably because there is not enough statistics. Results may improve somewhat once the initial probability tables are updated for the hybrid transforms which is coming soon. Change-Id: Ic7c0c62dacc68ef551054fdb575be8b8507d32a8
2012-09-10 07:42:35 +02:00
}
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
// horizontal transformation
pfa = &bufa[0];
pfb = &bufb[0];
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
switch (tx_type) {
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
case ADST_ADST :
case DCT_ADST :
hybrid transform of 16x16 dimension Enable ADST/DCT of dimension 16x16 for I16X16 modes. This change provides benefits mostly for hd sequences. Set up the framework for selectable transform dimension. Also allowing quantization parameter threshold to control the use of hybrid transform (This is currently disabled by setting threshold always above the quantization parameter. Adaptive thresholding can be built upon this, which will further improve the coding performance.) The coding performance gains (with respect to the codec that has all other configuration settings turned on) are derf: 0.013 yt: 0.086 hd: 0.198 std-hd: 0.501 Change-Id: Ibb4263a61fc74e0b3c345f54d73e8c73552bf926
2012-08-29 20:25:38 +02:00
pth = (tx_dim == 4) ? &adst_4[0] :
((tx_dim == 8) ? &adst_8[0] : &adst_16[0]);
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
break;
default :
hybrid transform of 16x16 dimension Enable ADST/DCT of dimension 16x16 for I16X16 modes. This change provides benefits mostly for hd sequences. Set up the framework for selectable transform dimension. Also allowing quantization parameter threshold to control the use of hybrid transform (This is currently disabled by setting threshold always above the quantization parameter. Adaptive thresholding can be built upon this, which will further improve the coding performance.) The coding performance gains (with respect to the codec that has all other configuration settings turned on) are derf: 0.013 yt: 0.086 hd: 0.198 std-hd: 0.501 Change-Id: Ibb4263a61fc74e0b3c345f54d73e8c73552bf926
2012-08-29 20:25:38 +02:00
pth = (tx_dim == 4) ? &dct_4[0] :
((tx_dim == 8) ? &dct_8[0] : &dct_16[0]);
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
break;
}
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
for (j = 0; j < tx_dim; j++) {
for (i = 0; i < tx_dim; i++) {
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfa[i] = 0;
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
for (k = 0; k < tx_dim; k++) {
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfa[i] += pfb[k] * pth[k];
}
pth += tx_dim;
}
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
pfa += tx_dim;
pfb += tx_dim;
// pth -= tx_dim * tx_dim;
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
switch (tx_type) {
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
case ADST_ADST :
case DCT_ADST :
pth = (tx_dim == 4) ? &adst_4[0] :
((tx_dim == 8) ? &adst_8[0] : &adst_16[0]);
break;
default :
pth = (tx_dim == 4) ? &dct_4[0] :
((tx_dim == 8) ? &dct_8[0] : &dct_16[0]);
break;
}
}
// convert to short integer format and load BLOCKD buffer
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
op = output;
pfa = &bufa[0];
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
for (j = 0; j < tx_dim; j++) {
for (i = 0; i < tx_dim; i++) {
op[i] = (pfa[i] > 0 ) ? (int16_t)( 8 * pfa[i] + 0.49) :
-(int16_t)(- 8 * pfa[i] + 0.49);
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
}
op += tx_dim;
pfa += tx_dim;
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
}
}
Remove vp8 in local symbols. For non-static functions, change the prefix to vp9_. For static functions, remove the prefix. Also fix some comments, remove unused code or unused function prototypes. Change-Id: I1f8be05362f66060fe421c3d4c9a906fdf835de5
2012-10-31 22:40:53 +01:00
vp9_clear_system_state(); // Make it simd safe : __asm emms;
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
}
/* Converted the transforms to integer form. */
#define VERTICAL_SHIFT 11
#define VERTICAL_ROUNDING ((1 << (VERTICAL_SHIFT - 1)) - 1)
#define HORIZONTAL_SHIFT 16
#define HORIZONTAL_ROUNDING ((1 << (HORIZONTAL_SHIFT - 1)) - 1)
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_fht_int_c(const int16_t *input, int pitch, int16_t *output,
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
TX_TYPE tx_type, int tx_dim) {
int i, j, k;
int16_t imbuf[256];
const int16_t *ip = input;
int16_t *op = output;
int16_t *im = &imbuf[0];
/* pointers to vertical and horizontal transforms. */
const int16_t *ptv = NULL, *pth = NULL;
switch (tx_type) {
case ADST_ADST :
ptv = pth = (tx_dim == 4) ? &adst_i4[0]
: ((tx_dim == 8) ? &adst_i8[0]
: &adst_i16[0]);
break;
case ADST_DCT :
ptv = (tx_dim == 4) ? &adst_i4[0]
: ((tx_dim == 8) ? &adst_i8[0] : &adst_i16[0]);
pth = (tx_dim == 4) ? &dct_i4[0]
: ((tx_dim == 8) ? &dct_i8[0] : &dct_i16[0]);
break;
case DCT_ADST :
ptv = (tx_dim == 4) ? &dct_i4[0]
: ((tx_dim == 8) ? &dct_i8[0] : &dct_i16[0]);
pth = (tx_dim == 4) ? &adst_i4[0]
: ((tx_dim == 8) ? &adst_i8[0] : &adst_i16[0]);
break;
case DCT_DCT :
ptv = pth = (tx_dim == 4) ? &dct_i4[0]
: ((tx_dim == 8) ? &dct_i8[0] : &dct_i16[0]);
break;
default:
assert(0);
break;
}
/* vertical transformation */
for (j = 0; j < tx_dim; j++) {
for (i = 0; i < tx_dim; i++) {
int temp = 0;
for (k = 0; k < tx_dim; k++) {
temp += ptv[k] * ip[(k * (pitch >> 1))];
}
im[i] = (int16_t)((temp + VERTICAL_ROUNDING) >> VERTICAL_SHIFT);
ip++;
}
Fixed style warnings Fixed some style warnings. Change-Id: I9eb77c3bf6d63628feef1b0c936e905bf4f2089d
2012-10-31 01:54:22 +01:00
im += tx_dim; // 16
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
ptv += tx_dim;
ip = input;
}
/* horizontal transformation */
im = &imbuf[0];
for (j = 0; j < tx_dim; j++) {
const int16_t *pthc = pth;
for (i = 0; i < tx_dim; i++) {
int temp = 0;
for (k = 0; k < tx_dim; k++) {
temp += im[k] * pthc[k];
}
op[i] = (int16_t)((temp + HORIZONTAL_ROUNDING) >> HORIZONTAL_SHIFT);
pthc += tx_dim;
}
Fixed style warnings Fixed some style warnings. Change-Id: I9eb77c3bf6d63628feef1b0c936e905bf4f2089d
2012-10-31 01:54:22 +01:00
im += tx_dim; // 16
Convert the transforms to integer forms. Converted the forward and inverse transforms to integer forms. Modify #define TEST_INT 1/0 in the code to call integer/float version of transforms. The tests showed that average OVERALL PSNR loss was less than 0.1%. Change-Id: I1dfa4eeab6412597e3b970ce299cf0e116a917e6
2012-10-19 01:31:59 +02:00
op += tx_dim;
}
Replacing the 8x8 DCT with 8x8 ADST/DCT for I8x8 Fixed the code review comments. Under the htrans8x8 experiment the 8X8 DCT in the I8X8 mode is replaced with a combination of 8X8 ADST and DCT. Overall coding gains with the htrans8x8 experiment are: derf: 0.486 std-hd: 1.040 hd: 1.063 yt: 0.506 Note that part of the gain comes from bigger transforms (8x8 instead of 4x4) and part comes from replacing the DCT wth the ADST. Change-Id: I92ca6bbfce11b4165d612b81d9adfad4d010c775
2012-08-02 18:07:33 +02:00
}
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_short_fdct4x4_c(short *input, short *output, int pitch) {
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
int i;
int a1, b1, c1, d1;
short *ip = input;
short *op = output;
Initial WebM release
2010-05-18 17:58:33 +02:00
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
for (i = 0; i < 4; i++) {
a1 = ((ip[0] + ip[3]) << 5);
b1 = ((ip[1] + ip[2]) << 5);
c1 = ((ip[1] - ip[2]) << 5);
d1 = ((ip[0] - ip[3]) << 5);
Merge Extended Q experiment. Merge the extended Q experiment as indicated by the Change-Id: I02d9e654fff9998cc7e9e2f1f5cd838dad8fb431
2012-02-09 17:44:46 +01:00
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
op[0] = a1 + b1;
op[2] = a1 - b1;
Initial WebM release
2010-05-18 17:58:33 +02:00
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
op[1] = (c1 * 2217 + d1 * 5352 + 14500) >> 12;
op[3] = (d1 * 2217 - c1 * 5352 + 7500) >> 12;
Initial WebM release
2010-05-18 17:58:33 +02:00
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
ip += pitch / 2;
op += 4;
Initial WebM release
2010-05-18 17:58:33 +02:00
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
}
ip = output;
op = output;
for (i = 0; i < 4; i++) {
a1 = ip[0] + ip[12];
b1 = ip[4] + ip[8];
c1 = ip[4] - ip[8];
d1 = ip[0] - ip[12];
op[0] = (a1 + b1 + 7) >> 4;
op[8] = (a1 - b1 + 7) >> 4;
op[4] = ((c1 * 2217 + d1 * 5352 + 12000) >> 16) + (d1 != 0);
op[12] = (d1 * 2217 - c1 * 5352 + 51000) >> 16;
ip++;
op++;
}
Initial WebM release
2010-05-18 17:58:33 +02:00
}
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_short_fdct8x4_c(short *input, short *output, int pitch)
Adds hybrid transform Adds ADST/DCT hybrid transform coding for Intra4x4 mode. The ADST is applied to directions in which the boundary pixels are used for prediction, while DCT applied to directions without corresponding boundary prediction. Adds enum TX_TYPE in b_mode_infor to indicate the transform type used. Make coding style consistent with google style. Fixed the commented issues. Experimental results in terms of bit-rate reduction: derf: 0.731% yt: 0.982% std-hd: 0.459% hd: 0.725% Will be looking at 8x8 transforms next. Change-Id: I46dbd7b80dbb3e8856e9c34fbc58cb3764a12fcf
2012-06-25 21:26:09 +02:00
{
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
vp9_short_fdct4x4_c(input, output, pitch);
vp9_short_fdct4x4_c(input + 4, output + 16, pitch);
fix more merge issues With this fix, the experimental branch now builds and encodes correctly with the following two configure options respectively: --enable-experimental --enable-t8x8 --enable-experimental Change-Id: I3147c33c503fe713a85fd371e4f1a974805778bf
2011-07-20 23:21:24 +02:00
}
Initial WebM release
2010-05-18 17:58:33 +02:00
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_short_walsh4x4_c(short *input, short *output, int pitch) {
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
int i;
int a1, b1, c1, d1;
short *ip = input;
short *op = output;
int pitch_short = pitch >> 1;
for (i = 0; i < 4; i++) {
a1 = ip[0 * pitch_short] + ip[3 * pitch_short];
b1 = ip[1 * pitch_short] + ip[2 * pitch_short];
c1 = ip[1 * pitch_short] - ip[2 * pitch_short];
d1 = ip[0 * pitch_short] - ip[3 * pitch_short];
op[0] = (a1 + b1 + 1) >> 1;
op[4] = (c1 + d1) >> 1;
op[8] = (a1 - b1) >> 1;
op[12] = (d1 - c1) >> 1;
ip++;
op++;
}
ip = output;
op = output;
for (i = 0; i < 4; i++) {
a1 = ip[0] + ip[3];
b1 = ip[1] + ip[2];
c1 = ip[1] - ip[2];
d1 = ip[0] - ip[3];
op[0] = (a1 + b1 + 1) >> 1;
op[1] = (c1 + d1) >> 1;
op[2] = (a1 - b1) >> 1;
op[3] = (d1 - c1) >> 1;
ip += 4;
op += 4;
}
Add lossless compression mode. This commit adds lossless compression capability to the experimental branch. The lossless experiment can be enabled using --enable-lossless in configure. When the experiment is enabled, the encoder will use lossless compression mode by command line option --lossless, and the decoder automatically recognizes a losslessly encoded clip and decodes accordingly. To achieve the lossless coding, this commit has changed the following: 1. To encode at lossless mode, encoder forces the use of unit quantizer, i.e, Q 0, where effective quantization is 1. Encoder also disables the usage of 8x8 transform and allows only 4x4 transform; 2. At Q 0, the first order 4x4 DCT/IDCT have been switched over to a pair of forward and inverse Walsh-Hadamard Transform (http://goo.gl/EIsfy), with proper scaling applied to match the range of the original 4x4 DCT/IDCT pair; 3. At Q 0, the second order remains to use the previous walsh-hadamard transform pair. However, to maintain the reversibility in second order transform at Q 0, scaling down is applied to first order DC coefficients prior to forward transform, and scaling up is applied to the second order output prior to quantization. Symmetric upscaling and downscaling are added around inverse second order transform; 4. At lossless mode, encoder also disables a number of minor features to ensure no loss is introduced, these features includes: a. Trellis quantization optimization b. Loop filtering c. Aggressive zero-binning, rounding and zero-bin boosting d. Mode based zero-bin boosting Lossless coding test was performed on all clips within the derf set, to verify that the commit has achieved lossless compression for all clips. The average compression ratio is around 2.57 to 1. (http://goo.gl/dEShs) Change-Id: Ia3aba7dd09df40dd590f93b9aba134defbc64e34
2012-06-14 04:03:31 +02:00
}
#if CONFIG_LOSSLESS
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_short_walsh4x4_lossless_c(short *input, short *output, int pitch) {
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
int i;
int a1, b1, c1, d1;
short *ip = input;
short *op = output;
int pitch_short = pitch >> 1;
for (i = 0; i < 4; i++) {
a1 = (ip[0 * pitch_short] + ip[3 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR;
b1 = (ip[1 * pitch_short] + ip[2 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR;
c1 = (ip[1 * pitch_short] - ip[2 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR;
d1 = (ip[0 * pitch_short] - ip[3 * pitch_short]) >> Y2_WHT_UPSCALE_FACTOR;
op[0] = (a1 + b1 + 1) >> 1;
op[4] = (c1 + d1) >> 1;
op[8] = (a1 - b1) >> 1;
op[12] = (d1 - c1) >> 1;
ip++;
op++;
}
ip = output;
op = output;
for (i = 0; i < 4; i++) {
a1 = ip[0] + ip[3];
b1 = ip[1] + ip[2];
c1 = ip[1] - ip[2];
d1 = ip[0] - ip[3];
op[0] = ((a1 + b1 + 1) >> 1) << Y2_WHT_UPSCALE_FACTOR;
op[1] = ((c1 + d1) >> 1) << Y2_WHT_UPSCALE_FACTOR;
op[2] = ((a1 - b1) >> 1) << Y2_WHT_UPSCALE_FACTOR;
op[3] = ((d1 - c1) >> 1) << Y2_WHT_UPSCALE_FACTOR;
ip += 4;
op += 4;
}
Add lossless compression mode. This commit adds lossless compression capability to the experimental branch. The lossless experiment can be enabled using --enable-lossless in configure. When the experiment is enabled, the encoder will use lossless compression mode by command line option --lossless, and the decoder automatically recognizes a losslessly encoded clip and decodes accordingly. To achieve the lossless coding, this commit has changed the following: 1. To encode at lossless mode, encoder forces the use of unit quantizer, i.e, Q 0, where effective quantization is 1. Encoder also disables the usage of 8x8 transform and allows only 4x4 transform; 2. At Q 0, the first order 4x4 DCT/IDCT have been switched over to a pair of forward and inverse Walsh-Hadamard Transform (http://goo.gl/EIsfy), with proper scaling applied to match the range of the original 4x4 DCT/IDCT pair; 3. At Q 0, the second order remains to use the previous walsh-hadamard transform pair. However, to maintain the reversibility in second order transform at Q 0, scaling down is applied to first order DC coefficients prior to forward transform, and scaling up is applied to the second order output prior to quantization. Symmetric upscaling and downscaling are added around inverse second order transform; 4. At lossless mode, encoder also disables a number of minor features to ensure no loss is introduced, these features includes: a. Trellis quantization optimization b. Loop filtering c. Aggressive zero-binning, rounding and zero-bin boosting d. Mode based zero-bin boosting Lossless coding test was performed on all clips within the derf set, to verify that the commit has achieved lossless compression for all clips. The average compression ratio is around 2.57 to 1. (http://goo.gl/dEShs) Change-Id: Ia3aba7dd09df40dd590f93b9aba134defbc64e34
2012-06-14 04:03:31 +02:00
}
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_short_walsh4x4_x8_c(short *input, short *output, int pitch) {
Restyle code Approximate the Google style guide[1] so that that there's a written document to follow and tools to check compliance[2]. [1]: http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml [2]: http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py Change-Id: Idf40e3d8dddcc72150f6af127b13e5dab838685f
2012-07-14 00:21:29 +02:00
int i;
int a1, b1, c1, d1;
short *ip = input;
short *op = output;
int pitch_short = pitch >> 1;
for (i = 0; i < 4; i++) {
a1 = ip[0 * pitch_short] + ip[3 * pitch_short];
b1 = ip[1 * pitch_short] + ip[2 * pitch_short];
c1 = ip[1 * pitch_short] - ip[2 * pitch_short];
d1 = ip[0 * pitch_short] - ip[3 * pitch_short];
op[0] = (a1 + b1 + 1) >> 1;
op[4] = (c1 + d1) >> 1;
op[8] = (a1 - b1) >> 1;
op[12] = (d1 - c1) >> 1;
ip++;
op++;
}
ip = output;
op = output;
for (i = 0; i < 4; i++) {
a1 = ip[0] + ip[3];
b1 = ip[1] + ip[2];
c1 = ip[1] - ip[2];
d1 = ip[0] - ip[3];
op[0] = ((a1 + b1 + 1) >> 1) << WHT_UPSCALE_FACTOR;
op[1] = ((c1 + d1) >> 1) << WHT_UPSCALE_FACTOR;
op[2] = ((a1 - b1) >> 1) << WHT_UPSCALE_FACTOR;
op[3] = ((d1 - c1) >> 1) << WHT_UPSCALE_FACTOR;
ip += 4;
op += 4;
}
Add lossless compression mode. This commit adds lossless compression capability to the experimental branch. The lossless experiment can be enabled using --enable-lossless in configure. When the experiment is enabled, the encoder will use lossless compression mode by command line option --lossless, and the decoder automatically recognizes a losslessly encoded clip and decodes accordingly. To achieve the lossless coding, this commit has changed the following: 1. To encode at lossless mode, encoder forces the use of unit quantizer, i.e, Q 0, where effective quantization is 1. Encoder also disables the usage of 8x8 transform and allows only 4x4 transform; 2. At Q 0, the first order 4x4 DCT/IDCT have been switched over to a pair of forward and inverse Walsh-Hadamard Transform (http://goo.gl/EIsfy), with proper scaling applied to match the range of the original 4x4 DCT/IDCT pair; 3. At Q 0, the second order remains to use the previous walsh-hadamard transform pair. However, to maintain the reversibility in second order transform at Q 0, scaling down is applied to first order DC coefficients prior to forward transform, and scaling up is applied to the second order output prior to quantization. Symmetric upscaling and downscaling are added around inverse second order transform; 4. At lossless mode, encoder also disables a number of minor features to ensure no loss is introduced, these features includes: a. Trellis quantization optimization b. Loop filtering c. Aggressive zero-binning, rounding and zero-bin boosting d. Mode based zero-bin boosting Lossless coding test was performed on all clips within the derf set, to verify that the commit has achieved lossless compression for all clips. The average compression ratio is around 2.57 to 1. (http://goo.gl/dEShs) Change-Id: Ia3aba7dd09df40dd590f93b9aba134defbc64e34
2012-06-14 04:03:31 +02:00
}
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_short_walsh8x4_x8_c(short *input, short *output, int pitch) {
vp9_short_walsh4x4_x8_c(input, output, pitch);
vp9_short_walsh4x4_x8_c(input + 4, output + 16, pitch);
Add lossless compression mode. This commit adds lossless compression capability to the experimental branch. The lossless experiment can be enabled using --enable-lossless in configure. When the experiment is enabled, the encoder will use lossless compression mode by command line option --lossless, and the decoder automatically recognizes a losslessly encoded clip and decodes accordingly. To achieve the lossless coding, this commit has changed the following: 1. To encode at lossless mode, encoder forces the use of unit quantizer, i.e, Q 0, where effective quantization is 1. Encoder also disables the usage of 8x8 transform and allows only 4x4 transform; 2. At Q 0, the first order 4x4 DCT/IDCT have been switched over to a pair of forward and inverse Walsh-Hadamard Transform (http://goo.gl/EIsfy), with proper scaling applied to match the range of the original 4x4 DCT/IDCT pair; 3. At Q 0, the second order remains to use the previous walsh-hadamard transform pair. However, to maintain the reversibility in second order transform at Q 0, scaling down is applied to first order DC coefficients prior to forward transform, and scaling up is applied to the second order output prior to quantization. Symmetric upscaling and downscaling are added around inverse second order transform; 4. At lossless mode, encoder also disables a number of minor features to ensure no loss is introduced, these features includes: a. Trellis quantization optimization b. Loop filtering c. Aggressive zero-binning, rounding and zero-bin boosting d. Mode based zero-bin boosting Lossless coding test was performed on all clips within the derf set, to verify that the commit has achieved lossless compression for all clips. The average compression ratio is around 2.57 to 1. (http://goo.gl/dEShs) Change-Id: Ia3aba7dd09df40dd590f93b9aba134defbc64e34
2012-06-14 04:03:31 +02:00
}
#endif
16x16 DCT blocks. Set on all 16x16 intra/inter modes Features: - Butterfly fDCT/iDCT - Loop filter does not filter internal edges with 16x16 - Optimize coefficient function - Update coefficient probability function - RD - Entropy stats - 16x16 is a config option Have not tested with experiments. hd: 2.60% std-hd: 2.43% yt: 1.32% derf: 0.60% Change-Id: I96fb090517c30c5da84bad4fae602c3ec0c58b1c
2012-08-03 02:03:14 +02:00
A bit of temporary relief on encoder slowness The commit replaces run-time initialization of cosine constants with static constant values, which provides ~30% relief on slow speed. The real solution, however will be to implement integer versions of those functions that current use float/double. Change-Id: Ie3ff1793509653d78dd1aeaf88cc6737da1bc55f
2012-08-07 22:55:49 +02:00
static const double C1 = 0.995184726672197;
static const double C2 = 0.98078528040323;
static const double C3 = 0.956940335732209;
static const double C4 = 0.923879532511287;
static const double C5 = 0.881921264348355;
static const double C6 = 0.831469612302545;
static const double C7 = 0.773010453362737;
static const double C8 = 0.707106781186548;
static const double C9 = 0.634393284163646;
static const double C10 = 0.555570233019602;
static const double C11 = 0.471396736825998;
static const double C12 = 0.38268343236509;
static const double C13 = 0.290284677254462;
static const double C14 = 0.195090322016128;
static const double C15 = 0.098017140329561;
16x16 DCT blocks. Set on all 16x16 intra/inter modes Features: - Butterfly fDCT/iDCT - Loop filter does not filter internal edges with 16x16 - Optimize coefficient function - Update coefficient probability function - RD - Entropy stats - 16x16 is a config option Have not tested with experiments. hd: 2.60% std-hd: 2.43% yt: 1.32% derf: 0.60% Change-Id: I96fb090517c30c5da84bad4fae602c3ec0c58b1c
2012-08-03 02:03:14 +02:00
static void dct16x16_1d(double input[16], double output[16]) {
Remove vp8 in local symbols. For non-static functions, change the prefix to vp9_. For static functions, remove the prefix. Also fix some comments, remove unused code or unused function prototypes. Change-Id: I1f8be05362f66060fe421c3d4c9a906fdf835de5
2012-10-31 22:40:53 +01:00
vp9_clear_system_state(); // Make it simd safe : __asm emms;
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
{
double step[16];
double intermediate[16];
double temp1, temp2;
// step 1
step[ 0] = input[0] + input[15];
step[ 1] = input[1] + input[14];
step[ 2] = input[2] + input[13];
step[ 3] = input[3] + input[12];
step[ 4] = input[4] + input[11];
step[ 5] = input[5] + input[10];
step[ 6] = input[6] + input[ 9];
step[ 7] = input[7] + input[ 8];
step[ 8] = input[7] - input[ 8];
step[ 9] = input[6] - input[ 9];
step[10] = input[5] - input[10];
step[11] = input[4] - input[11];
step[12] = input[3] - input[12];
step[13] = input[2] - input[13];
step[14] = input[1] - input[14];
step[15] = input[0] - input[15];
// step 2
output[0] = step[0] + step[7];
output[1] = step[1] + step[6];
output[2] = step[2] + step[5];
output[3] = step[3] + step[4];
output[4] = step[3] - step[4];
output[5] = step[2] - step[5];
output[6] = step[1] - step[6];
output[7] = step[0] - step[7];
temp1 = step[ 8]*C7;
temp2 = step[15]*C9;
output[ 8] = temp1 + temp2;
temp1 = step[ 9]*C11;
temp2 = step[14]*C5;
output[ 9] = temp1 - temp2;
temp1 = step[10]*C3;
temp2 = step[13]*C13;
output[10] = temp1 + temp2;
temp1 = step[11]*C15;
temp2 = step[12]*C1;
output[11] = temp1 - temp2;
temp1 = step[11]*C1;
temp2 = step[12]*C15;
output[12] = temp2 + temp1;
temp1 = step[10]*C13;
temp2 = step[13]*C3;
output[13] = temp2 - temp1;
temp1 = step[ 9]*C5;
temp2 = step[14]*C11;
output[14] = temp2 + temp1;
temp1 = step[ 8]*C9;
temp2 = step[15]*C7;
output[15] = temp2 - temp1;
// step 3
step[ 0] = output[0] + output[3];
step[ 1] = output[1] + output[2];
step[ 2] = output[1] - output[2];
step[ 3] = output[0] - output[3];
temp1 = output[4]*C14;
temp2 = output[7]*C2;
step[ 4] = temp1 + temp2;
temp1 = output[5]*C10;
temp2 = output[6]*C6;
step[ 5] = temp1 + temp2;
temp1 = output[5]*C6;
temp2 = output[6]*C10;
step[ 6] = temp2 - temp1;
temp1 = output[4]*C2;
temp2 = output[7]*C14;
step[ 7] = temp2 - temp1;
step[ 8] = output[ 8] + output[11];
step[ 9] = output[ 9] + output[10];
step[10] = output[ 9] - output[10];
step[11] = output[ 8] - output[11];
step[12] = output[12] + output[15];
step[13] = output[13] + output[14];
step[14] = output[13] - output[14];
step[15] = output[12] - output[15];
// step 4
output[ 0] = (step[ 0] + step[ 1]);
output[ 8] = (step[ 0] - step[ 1]);
temp1 = step[2]*C12;
temp2 = step[3]*C4;
temp1 = temp1 + temp2;
output[ 4] = 2*(temp1*C8);
temp1 = step[2]*C4;
temp2 = step[3]*C12;
temp1 = temp2 - temp1;
output[12] = 2*(temp1*C8);
output[ 2] = 2*((step[4] + step[ 5])*C8);
output[14] = 2*((step[7] - step[ 6])*C8);
temp1 = step[4] - step[5];
temp2 = step[6] + step[7];
output[ 6] = (temp1 + temp2);
output[10] = (temp1 - temp2);
intermediate[8] = step[8] + step[14];
intermediate[9] = step[9] + step[15];
temp1 = intermediate[8]*C12;
temp2 = intermediate[9]*C4;
temp1 = temp1 - temp2;
output[3] = 2*(temp1*C8);
temp1 = intermediate[8]*C4;
temp2 = intermediate[9]*C12;
temp1 = temp2 + temp1;
output[13] = 2*(temp1*C8);
output[ 9] = 2*((step[10] + step[11])*C8);
intermediate[11] = step[10] - step[11];
intermediate[12] = step[12] + step[13];
intermediate[13] = step[12] - step[13];
intermediate[14] = step[ 8] - step[14];
intermediate[15] = step[ 9] - step[15];
output[15] = (intermediate[11] + intermediate[12]);
output[ 1] = -(intermediate[11] - intermediate[12]);
output[ 7] = 2*(intermediate[13]*C8);
temp1 = intermediate[14]*C12;
temp2 = intermediate[15]*C4;
temp1 = temp1 - temp2;
output[11] = -2*(temp1*C8);
temp1 = intermediate[14]*C4;
temp2 = intermediate[15]*C12;
temp1 = temp2 + temp1;
output[ 5] = 2*(temp1*C8);
}
Remove vp8 in local symbols. For non-static functions, change the prefix to vp9_. For static functions, remove the prefix. Also fix some comments, remove unused code or unused function prototypes. Change-Id: I1f8be05362f66060fe421c3d4c9a906fdf835de5
2012-10-31 22:40:53 +01:00
vp9_clear_system_state(); // Make it simd safe : __asm emms;
16x16 DCT blocks. Set on all 16x16 intra/inter modes Features: - Butterfly fDCT/iDCT - Loop filter does not filter internal edges with 16x16 - Optimize coefficient function - Update coefficient probability function - RD - Entropy stats - 16x16 is a config option Have not tested with experiments. hd: 2.60% std-hd: 2.43% yt: 1.32% derf: 0.60% Change-Id: I96fb090517c30c5da84bad4fae602c3ec0c58b1c
2012-08-03 02:03:14 +02:00
}
Change encoder vp8_ and vp8cx_ public symbol prefixes to vp9_. Change-Id: Ie2e3652591b010ded10c216501ce24fd95d0aec5
2012-10-30 20:58:42 +01:00
void vp9_short_fdct16x16_c(short *input, short *out, int pitch) {
Remove vp8 in local symbols. For non-static functions, change the prefix to vp9_. For static functions, remove the prefix. Also fix some comments, remove unused code or unused function prototypes. Change-Id: I1f8be05362f66060fe421c3d4c9a906fdf835de5
2012-10-31 22:40:53 +01:00
vp9_clear_system_state(); // Make it simd safe : __asm emms;
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
{
16x16 DCT blocks. Set on all 16x16 intra/inter modes Features: - Butterfly fDCT/iDCT - Loop filter does not filter internal edges with 16x16 - Optimize coefficient function - Update coefficient probability function - RD - Entropy stats - 16x16 is a config option Have not tested with experiments. hd: 2.60% std-hd: 2.43% yt: 1.32% derf: 0.60% Change-Id: I96fb090517c30c5da84bad4fae602c3ec0c58b1c
2012-08-03 02:03:14 +02:00
int shortpitch = pitch >> 1;
int i, j;
double output[256];
// First transform columns
for (i = 0; i < 16; i++) {
double temp_in[16], temp_out[16];
for (j = 0; j < 16; j++)
temp_in[j] = input[j*shortpitch + i];
dct16x16_1d(temp_in, temp_out);
for (j = 0; j < 16; j++)
output[j*16 + i] = temp_out[j];
}
// Then transform rows
for (i = 0; i < 16; ++i) {
double temp_in[16], temp_out[16];
for (j = 0; j < 16; ++j)
temp_in[j] = output[j + i*16];
dct16x16_1d(temp_in, temp_out);
for (j = 0; j < 16; ++j)
output[j + i*16] = temp_out[j];
}
// Scale by some magic number
for (i = 0; i < 256; i++)
out[i] = (short)round(output[i]/2);
Fix SIMD unsafe use of floating point. This commit fixes unsafe simd / floating point interactions arising from the current hybrid and 16x16 transform implementation. These led to a raft of bugs and issues when the project was built using VS2008 for Win32 though they did not show up with the unix builds. Gerrit makes a meal out of presenting the fix but all I have actually done is indent the body of each function that uses floating point by one level and bracket with emms instructions using the function vp8_clear_system_state(). See below. function () { vp8_clear_system_state(); { ... function body } vp8_clear_system_state(); } This is almost certainly over the top in terms of number of emms instructions but is a temporary measure pending implementation of integer variants of each function to replace the floating point. Limited testing suggests that this fixes the problems that arose for Win32 VS2008 when the hybrid or 16x16 transforms were enabled. Change-Id: I7c9a72bd79315246ed880578dec51e2b7c178442
2012-10-05 12:16:46 +02:00
}
Remove vp8 in local symbols. For non-static functions, change the prefix to vp9_. For static functions, remove the prefix. Also fix some comments, remove unused code or unused function prototypes. Change-Id: I1f8be05362f66060fe421c3d4c9a906fdf835de5
2012-10-31 22:40:53 +01:00
vp9_clear_system_state(); // Make it simd safe : __asm emms;
16x16 DCT blocks. Set on all 16x16 intra/inter modes Features: - Butterfly fDCT/iDCT - Loop filter does not filter internal edges with 16x16 - Optimize coefficient function - Update coefficient probability function - RD - Entropy stats - 16x16 is a config option Have not tested with experiments. hd: 2.60% std-hd: 2.43% yt: 1.32% derf: 0.60% Change-Id: I96fb090517c30c5da84bad4fae602c3ec0c58b1c
2012-08-03 02:03:14 +02:00
}
Reference in New Issue Copy Permalink